Enhancing Energy System Resilience Through Advanced Detection of Extreme Weather Events

Within the EnergyProtect project the escalating risks posed by extreme weather events to renewable energy infrastructure and affects on production are tackled. As climate change intensifies storms, heatwaves, and heavy precipitation, the vulnerability of renewable energy systems demands advanced detection and prediction methods to ensure resilience. The project focuses on developing machine learning algorithms for detecting adverse weather patterns, as well as dynamical and machine learning physics-informed downscaling, to enable precise risk assessment and infrastructure protection strategies tailored to current and future conditions.

Central to EnergyProtect is a three-tiered methodology for weather risk detection and resilience assessment:

• Machine Learning detection methods: Advanced pattern detection algorithms integrate atmospheric domain knowledge to identify and classify high-risk weather patterns. These models improve detection accuracy by combining meteorological parameters with infrastructure-specific indicators.
• High-Resolution Physics-Aware and dynamical Downscaling: Convection-permitting models at 1–2 km resolution enable detailed simulations of localized extreme weather events, addressing the challenges of complex terrain where traditional models often fail.
• Probabilistic Risk Assessment: Infrastructure vulnerability data is combined with detected weather patterns to quantify resilience under various scenarios, incorporating economic incentives and regulatory frameworks to support adaptation.

Here, we show initial results of the adverse atmspheric event detection algrithms threatening energy infrastructure. Different methods, classical weather pattern and machine learning algorithms, are investigated. Historic events such as the storm Boris and events defined with the industry stakeholders are evaluated.  Additinally, a first set of dynamical downscaled climate scenarios is used for selected adverse weather types to evalute the methods skills across the different resolution scales.